Feature Selection and Weighted Based Optimized Weight based Multi-Tier Stacked Ensemble (WMTSE) Classification for Twitter Sentiment Analysis

Abstract

Abstract This research work concentrates on both feature selection and classification methods for utilizing twitter data. A new classifier is introduced for classifying “tweets” into positive, negative and neutral sentiment. The system contains four steps: Preprocessing by Tokenization, Text Cleaning, Part of Speech (PoS) Tagging, Stemming and Stop Words Removal, Feature Extraction by Bag-of-words (BoW), Lexicon-based features and Term Frequency- Inverse Document Frequency(TF-IDF), Feature Selection by Binary Swallow Swarm Optimization (BSSO) and the Classification Model by Weighting Naive Bayes (WNB), Multi-Tier Stacked Ensemble (MTSE) and Optimized Weight based Multi-Tier Stacked Ensemble (OWMTSE) for Sentiment Analysis. These methods are implemented via the collected dataset of sentiment140 dataset. It contains 1,600,000 tweets extracted using the twitter API. The tweets have been annotated (0 = negative, 2 = neutral, 4 = positive) and they can be used to detect sentiment [6]. Keywords: Twitter, sentiment, Bag of words, BSSO, MTSE, OWMTSE